| In the power system,power transformer is the core of the power transmission and transformation equipment,connected with a plurality of voltage levels,is the hub of the power grid.According to the power grid and China Southern Power Grid National Statistics: in 2004 the national grid above 110 kV power transformer winding damage according to the site statistics show 37 times,accounting for 69.8% of the total;Guangxi power grid in 2013 5 times the total damage,winding damage times for 4 times,accounting for 80%.Most of the damage is not winding insulation,but the lack of mechanical performance caused by electrical faults,the impact on the stable operation of the transformer.The winding deformation has the characteristics of concealment and accumulation,so it is of great significance to realize the on-line diagnosis of winding deformation.The off-line test data show that the impact has little effect on core vibration,after three rounds of short-circuit,the vibration signal frequency distribution changes little,the standard deviation of the amplitude frequency were 0.0711,0.0778,0.0775.In this paper,the frequency band energy Euclidean distance obtained by wavelet packet decomposition is used to calculate the deformation of the winding.Taking A as an example,the characteristic values of the three phase impact test were 0.0648,0.1094,and 0.2062,respectively.In this paper,the blind source separation algorithm is used to separate the vibration signal of the transformer on line.The simulation results and the analysis of the network data show that the algorithm can be used to separate the vibration signal of the transformer.Neural network is used to train the network to achieve the purpose of identifying the vibration signal of the winding.For three 220 kV operation of historical statistics,comparative analysis of the vibration signal under the same load,verify the "energy band-Euclidean distance as the vibration characteristics is also suitable for 220 kV transformer,just the diagnostic threshold and test of transformer.There are a lot of deficiencies in the result of single data diagnosis,and the reliability and accuracy is not high.Based on the realization method of transformer winding deformation vibration detection,this paper through the analysis of short-circuit reactance and vibration characteristics of the multi information models for diagnosing transformer winding deformation based on the results of data analysis indicated that it had a higher accuracy,smaller deformation also have the ability to distinguish. |
